Pachyderm: Easy Data Retrieval

INTRODUCTION

EASY DATA
RETRIEVAL

Most email systems have mechanisms that allow users to categorize their received email, in order to assist them in finding it later. Pachyderm allows this by letting users attached "labels" (simple strings) to messages and then letting them retrieve messages having particular labels. But in Pachyderm this is by no means the primary mechanism for organizing mail.

The only mechanism for retrieving messages from Pachyderm is by issuing a "query". The result of the query is a list of the matching messages, and the user can then select and display, reply to, or forward one or more of those messages.

The queries supported by Pachyderm are basically the same as the AltaVista web index's advanced query language. Pachyderm maintains a full text index of each user's email, augmented by any labels that the user has attached, and including structural information about each message such as its "from" or "subject" field or its date. The queries that the user issues are matched against this index.

The effect is dramatic. Users who have been collecting old email for many years might accumulate as many as 30,000 messages. This is a factor of 1,000 less than the number of web pages indexed by AltaVista, and the number of hits for a simple query seem to be correspondingly reduced. A query with just one or two terms will often produce only 10 hits, making it very easy for the user to select the desired message.

The query mechanism also subsumes other features found in earlier email systems. For example, instead of sorting email by sender, users just issue a query matching words in the "from" field. Instead of sorting email by subject, users just issue a specialized query that matches over "subject" and "in-reply-to" message ID fields.

Overall, this use of a query system seems to work much better than folder-based schemes. Users need make very little effort at classifying their email, and yet can still reliably find old messages.